1002 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 65, NO. 3, MARCH 2018

The Calculation of Negative Bias IlluminationStress-Induced Instability of Amorphous

InGaZnO Thin-Film Transistors forInstability-Aware Design

Jun Tae Jang, Sung-Jin Choi , Dong Myong Kim , Member, IEEE, andDae Hwan Kim , Senior Member, IEEE

Abstract— We propose a systematic calculation methodfor negative bias illumination stress (NBIS)-induced insta-bilities in amorphous InGaZnO thin-film transistors (TFTs).The proposed method is based on activation energy win-dow (AEW) in subgap energy range, and it can reproduce theNBIS time evolution of I–V characteristics without long-termstress test. Furthermore, it quantitatively explains the effectof oxygen content on the NBIS instability. The AEW, whichis employed for emulating the oxygen vacancy ionization,peroxide formation, and hole trapping, has the order of mag-nitude of 1016–1017 cm−3 for the bias stress of −20 V andthe commercial LED backlight of 300 cd/m2. The proposedmethod is expected to play such an important role in theinstability awareness of amorphous oxide TFTs.

Index Terms— Activation energy window (AEW),InGaZnO, instability, negative bias illumination stress(NBIS), thin-film transistors (TFTs).

I. INTRODUCTION

AMORPHOUS indium-gallium-zinc-oxide (a-IGZO)thin-film transistors (TFTs) have attracted much

attention due to its advantages, such as high field-effectmobility (μFE), low subthreshold swing (SS), high ON

current, and low OFF current, as promising candidatesfor switching or driving devices in the field of flat paneldisplays [1]–[3]. However, various instabilities underbias/illumination stress remain as challenging issues in termsof commercialization [2], [3]. To date, many researchers havequalitatively studied on mechanisms of the negative bias

Manuscript received October 11, 2017; revised December 11, 2017;accepted January 13, 2018. Date of publication February 5, 2018;date of current version February 22, 2018. This work was supportedin part by the National Research Foundation of Korea funded by theKorean Government (MSIP) under Grant 2016R1A5A1012966 and inpart by the Ministry of Education, Science and Technology underGrant 2017R1A2B4006982. The review of this paper was arranged byEditor K. C. Choi. (Corresponding author: Dae Hwan Kim.)

The authors are with the School of Electrical Engineering, KookminUniversity, Seoul 136-702, South Korea (e-mail: drlife@kookmin.ac.kr).

Color versions of one or more of the figures in this paper are availableonline at http://ieeexplore.ieee.org.

Digital Object Identifier 10.1109/TED.2018.2797208

illumination stress (NBIS) instability in a-IGZO TFTs, whichhave been attributed to the positively charged trapping modelonly with the negative shift of threshold voltage (VT ) [4], [5],and to the oxygen vacancy (VO) ionization model leading toboth the negative shift of the VT and degradation of the μFEas well as the SS [6], [7]. Recently, Nahm et al. [8] haveproposed a metastable peroxide formation model leadingto the negative shift of the VT to explain the NBIS-basedinstability in relation to persistent photoconductivity [8].However, the quantitative analysis on the NBIS-induced VT

shift (�VT ) has not been thoroughly studied, especially inperspective of applying physical long-term instability modelsto the instability-aware design of display circuits.

In this paper, a calculation method for NBIS instability ina-IGZO TFTs is proposed and demonstrated with two kindsof cases with oxygen contents in active layers, i.e., oxygen-poor (O-poor) and oxygen-rich (O-rich) devices. The proposedmethod effectively enables reproducing the NBIS time (tNBIS)evolution of current–voltage (I–V ) characteristics only fromthe pristine density-of-subgap states (DOS) and NBIS condi-tion even without long-term stress test and is very applicableto general amorphous oxide TFTs along with the process-/material-dependent DOS parameters.

II. CALCULATION METHOD FOR THE NBIS �VT

The NBIS instability mechanism of a-IGZO TFT can beclassified into following three types.

1) The optical illumination-generated holes or positivecharged species followed by their trapping into gateinsulator/interface as shown in Fig. 1(a), i.e., (hν) →e− + h+ with h = Plank’s constant and ν = frequencyof photon [4], [5].

2) The donor creation by oxygen vacancy ionization asdepicted in Fig. 1(b), i.e., VO + (hν) → V 2+

O +2e− or V 1+

O + 1e− [6], [7].3) The donor creation by the hole-mediated formation of

the metastable peroxide defects as described in Fig. 1(c),i.e., O2− + O2− + (hν) → O2−

2 + 2e− [8].

0018-9383 © 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

JANG et al.: CALCULATION OF NBIS-INDUCED INSTABILITY 1003

Fig. 1. Schematic illustrations of (a) charge trapping, (b) oxygenvacancy ionization, and (c) peroxide formation models. Schematic of(d) NBIS-induced instability mechanisms correlated with the DOS andAEW, (e) EBD under NBIS, and (f) method for calculating the AEW alonga channel depth direction.

The interaction between oxide TFT and ambient [9] wasexcluded from possible mechanisms due to the presence ofrobust passivation layer [10].

The valence band tail states (VBTs) of a-IGZO are locatedin the levels under mid-gap and consist of the (intrinsic) ppσ ∗antibonding defects [8] and the (extrinsic) oxygen vacancy-related defects [7], [11]. When the device is under a negativegate bias stress, the electron quasi-Fermi energy level EFnis located in VBTs. Therefore, the NBIS-induced �VT is

determined by how many subgap states exist in the energyrange between EFn and EC–Eph, i.e., activation energy win-dow (AEW) shown in Fig. 1(d), regardless of what thedominant instability mechanism is. Here, EC , EV , and Eph aredefined to be the conduction band minimum level, the valenceband maximum level, and the photon energy.

Fig. 1(e) and (f) shows the energy band diagrams (EBDs)under the NBIS. Here, the density of VBTs is symbolized asgD(E) [eV−1cm−3], which will be shown in Fig. 1(d). Then,AEW can be derived from the following equation:

AEW =∫ Tact

Tact−Teff

∫ EFnEC −Eph

gD(E)d Edx

Teff(1)

where x , Tact, and Teff are the position coordinate alongthe direction of TFT channel depth, the thickness of TFTactive layer, and the effective thickness of the AEW. Here,Teff means the effective thickness of IGZO film where thedegradation occurs under NBIS and it varies with the increaseof tNBIS as well as by changing the device structure, stressbias, photoillumination condition, and active material becauseit manifests itself that the AEW plays as a function ofthe process/material parameters via a pristine gD(E), NBIScondition via Eph, and the geometry of TFTs via Tact. TheEFn and Teff are then determined by specific conditions ofgD(E), Eph, and Tact.

The calculated AEW can be transformed into increasingthe concentration of shallow donor (NSD) by �NSD andmakes it possible to estimate the tNBIS evolution of TFT I–Vcharacteristics as follows.

If the AEW at t = t0 (pristine state), i.e., before applyingNBIS, is an AEW(t0), the �VT during �t under NBIS is deter-mined by �NSD(t = t0 + �t). In other word, the AEW(t0)is identical to the increase of shallow donor concentration inactive layer (�NSD) between adjacent time spots t = t0 andt = t0 + �t as follows:

AEW(t0) = �NSD(t = t0 + �t)

= NSD(t = t0 + �t) − NSD(t = t0). (2)

Therefore, the tNBIS evolution of AEW(t) can be calcu-lated as

AEW(ti )

=∫ Tact

Tact−Teff(ti )

∫ EFn(ti )EC −Eph

gD(E)d Edx

Teff(ti )= �NSD(t = ti + �t)= NSD(t = ti+1) − NSD(t = t0) (3)

AEW(ti+1) − AEW(ti )

= NSD(ti+2) − NSD(ti+1)

=∫ Tact

Tact−Teff(ti+1)

∫ EFn(ti+1)EC −Eph

gD(E)d Edx

Teff(ti+1)

−∫ Tact

Tact−Teff(ti )

∫ EFn(ti )EC −Eph

gD(E)d Edx

Teff(ti )

≈∫ Tact

Tact−Teff(ti )

∫ EFn(ti+1)EFn(ti )

gD(E)d Edx

Teff(ti )(4)

where ti is a specific time spot under NBIS and ti+1 is ti +�t .

1004 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 65, NO. 3, MARCH 2018

Fig. 2. (a) Schematic illustrations of fabricated a-IGZO TFT. (b) Transfercharacteristics with oxygen gas flow rate.

Fig. 3. (a) Measured C–V characteristics at dark ambient. The measuredphotoresponses of C–V characteristics in (b) O-poor and (c) O-rich TFTs.Extracted DOSs (d) in a logarithm and (e) linear scale. The oxygen-content-dependent (f) VT and (g) µFE,lin extracted at IDS = 1 nA andVGS = 15 V/VDS = 0.1 V from the measurement for six samples of eachoxygen gas flow rate condition.

Here, it should be noted that one can calculate all ofEFn(ti ), AEW(ti ), VT , and I–V characteristics from gD(E)and NSD(t = ti ) at a specific time spot by using our previouslyreported I–V model and TCAD simulator [12], [13]. There-fore, we can also calculate either the area of AEW or the I–Vcharacteristic as a function of tNBIS by iterating numericallythis procedure with an infinitesimal �t .

III. EXPERIMENTAL RESULTS AND DISCUSSION

The a-IGZO TFT devices have the inverted staggeredbottom-gate structure with an etch-stop layer as shownin Fig. 2(a). The sputter-deposited molybdenum (Mo) was

used as a metal for gate, source, and drain electrodes.A 400-nm-thick SiOx layer and a 50-nm-thick SiNx layerwere deposited by plasma-enhanced chemical vapor deposi-tion (PECVD) forming the gate insulator. Then, in order tocontrol the oxygen contents in active layers, a 50-nm-thickIGZO was deposited as TFT active layer by dc sputteringwith the gas flow rate of Ar/O2 = 35/21 sccm (O-poor)and 35/63 sccm (O-rich) at a total pressure of 5 mtorr.Moreover, the sputtering power was controlled to be 2 kWat room temperature. A 50-nm-thick SiOx etch-stop layer wasdeposited by PECVD and patterned by wet etching. For theformation of source/drain (S/D) electrodes, Mo is sputteredat room temperature and patterned by dry etching. Finally,the fabricated devices were thermally annealed at 250 °Cfor 1 h.

The channel width (W ) and length (L) of the deviceswere 100 and 100 μm, respectively. In addition, all I–V andcapacitance–voltage (C–V ) characteristics were measured atroom temperature through the Agilent 4156C semiconductingparameter analyzer and the Agilent 4284A precision LCRmeter, respectively. The C–V curve was measured betweenthe gate and the S/D tied each other as a function of gatevoltage (VG). The measured transfer characteristics with dif-ferent oxygen gas flow rates are depicted as shown in Fig. 2(b).

Fig. 3(a) shows the effect of oxygen content on C–V curvemeasured at a dark ambient, which is attributed to that eitherthe density of subgap traps or the carrier concentration varieswith the change of oxygen contents. First, the VOs play therole of shallow donor in IGZO. Thus, the O-poor device hashigher electron concentration, hence a lower VT . Therefore,the VG point where the capacitance begins to sharply riseshifts to more negative direction in O-poor device comparedwith O-rich device. Second, the VG sweeps from negative topositive voltage, which is equivalent with the movement ofEFn from EV up to EC . If a larger number of subgap statesexist below EC , the EFn is harder to lift up to EC and the slopeof dC/dV G becomes smaller. Therefore, the value of dC/dV G

is smaller in O-rich device [higher gA(E)] than that in O-poordevice [lower gA(E)] as will be discussed in Fig. 3(d) and (e).

In order to calculate the AEW, we then extracted DOSbased on the photoresponse of TFT C–V characteristics. Thelatter technique is referred as monochromatic photonic C–Vspectroscopy [14], and the I–V characteristics were calculatedby DeAOTS [12]. Fig. 3(b) and (c) shows the measured pho-toresponses of C–V characteristics, the conditions of whichare as follows: small-signal frequency of 100 kHz, illuminationwavelength of 532 nm, and optical power of 5 mW. The DOSsof two devices (O-poor and O-rich) are extracted as indicatedin Fig. 3(d) and (e) from the photoresponses of C–V shownin Fig. 3(b) and (c). It is found that the total DOS [g(E)]consists of the donor-like DOS [gD(E)] near EV and theacceptor-like DOS [gA(E)] near EC . The related details ofextraction can also be found in [14].

The material and DOS parameters which were used inour I–V simulation are summarized in Table I. The bandgap(Eg) was taken from spectroscopic ellipsometry. The valueof effective DOS (NC ) is calculated by consideration ofelectron effective mass (∼0.35 m0), and the value of NV is

JANG et al.: CALCULATION OF NBIS-INDUCED INSTABILITY 1005

TABLE IMATERIAL AND DOS PARAMETERS MEASURED FROM THE DEVICES

calculated from NC × (NTD/NTA) [12]. The value of NSD istaken by adjusting its value with numerical iteration until thecalculated I–V characteristics agree well with the measuredones. Used I–V model was given in [12] and [13]. All ofthe DOS parameters are taken from fitting the model equation[line in Fig. 3(d) and Table I] with experimentally extractedDOS [symbol in Fig. 3(d)].

It is found in Fig. 3(d) and (e) that the extracted gD(E) ofO-poor IGZO TFT is higher than that of the O-rich device. Thedifference of gD(E) between the O-poor and O-rich devices isprominent, especially in the level around EV +1 eV. It suggeststhat a larger amount of neutral oxygen vacancy defects existsin the O-poor TFT in comparison with O-rich TFT [11].Moreover, it was found that the gD(E) near EV is correlatedwith neutral oxygen vacancy defect through a dependence ofphotoconductivity characteristics with the oxygen gas flowrate in [15]. On the other hand, the density of acceptor-like traps near EC , i.e., gA(E), is higher in O-rich ratherthan in O-poor TFTs, which is attributed to more severeion bombardment (higher total flow rate of gas) during thedc sputtering [16]. It is consistent that in comparison withO-rich device, the photoresponse of O-poor device is largerdue to the higher gD(E) as depicted in C–V characteristicsof Fig. 3(b) and (c).

From the measurements for six samples shown as thesymbols in Fig. 3(f) and (g), it is observed that the O-poordevice has the definitely lower VT and higher μFE,lin than theO-rich device, where the VT and μFE,lin (μFE extracted in alinear region) are extracted at the IDS = 1 nA and VGS = 15V/VDS = 0.1 V, respectively. Lower VT and higher μFE,lin ofthe O-poor device are well correlated with larger amount ofVOs and lower gA(E) than the O-rich device.

Noticeably, the extracted DOS is validated by comparingDOS-based simulated I–V characteristics with the measuredpristine I–V characteristics [the black lines and black symbolsshown in Fig. 4(a) and (c)].

Fig. 4. Measured and calculated tNBIS dependence of I–V char-acteristics of (a) and (b) O-poor and (c) and (d) O-rich TFTs where(a) and (c) VDS = 0.1 V and (b) and (d) VDS = 10 V. The tNBIS-dependent(e) VT and (f) µFE,lin from I–V characteristics under VGS = 15 V andVDS = 0.1 V.

The measured tNBIS evolution of I–V characteristics ofa-IGZO TFTs are shown as symbols in Fig. 4(a)–(d), wherethe conditions of NBIS are VGS = −20 V, VDS = 10 V, illu-mination by a commercial LED backlight unit with brightnessof 300 cd/m2. Also, the measured tNBIS evolutions of VT andμFE,lin are shown as the symbols in Fig. 4(e) and (f). While theμE,lin remains nearly unchanged during NBIS, the magnitudeof NBIS-induced �VT is much smaller in O-rich device thanin O-poor device, which is consistent with [17] and [18].

In order to verify our calculation method, the tNBIS evo-lutions of I–V characteristics were calculated by using theDOS/AEW-based model which is proposed in Section II.The parameters used in calculating the NBIS instability aresummarized in Table II, where the EFB, VFB, μBand, andEph,eff are the value of difference between EC and EFn at aflat band condition, flat band voltage, conduction band electronmobility, and effective photon energy, respectively [12]. Asdenoted by the lines in Fig. 4(a)–(d), the measured tNBIS evolu-tions of I–V characteristics are reproduced very well by usingthe proposed method in O-rich as well as in O-poor device,which suggests that the extracted DOS and the calculationmethod proposed are reasonable.

1006 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 65, NO. 3, MARCH 2018

TABLE IIPARAMETERS USED IN CALCULATING THE NBIS INSTABILITY

The calculated tNBIS dependences of EFn and EC –Eph levelsare shown as the EBDs in Fig. 5(a) and (b), where the tNBISis denoted by the legends. It is found that in the case ofO-rich devices, the variation of EC–Eph levels during NBISis larger than O-poor devices, which can be explained asfollows. The gD(E) is higher in O-poor device rather thanin O-rich device. It means that at the same bias condition,the localized charge density is lower in O-rich device thanin O-poor device. In turn, the free carrier charge densityshould be higher in O-rich device. Therefore, in O-rich device,the EC is modulated larger either by the same change of gatevoltage or during the same NBIS time.

Thus, as shown in Fig. 5(a) and (b), the variation of EC –Ephlevel is larger, i.e., the band bending is more prominent duringNBIS, in O-rich device rather than in O-poor device. Then,the area of AEW is smaller in O-rich device than in O-poordevice. Consistently, the calculated tNBIS evolutions of Teffand AEW are shown in Fig. 5(c) and (d). An AEW, whichis much larger than that of the O-rich devices, in the O-poordevice, clearly explains the reason of the O-poor device beingless stable under the NBIS than the O-rich device.

In terms of the validity of AEW calculation, it shouldbe noted that the proposed method is available only onthe condition that the number of incident photons is largerthan the number of subgap states in AEW. In our case,the number of incident photons from commercial backlightunit was 1.4 × 1015 cm−2s−1. On the other hand, the numberof subgap states in AEW was 1.3×1011 cm−2 (O-poor device)and 7.5 × 1010 cm−2 (O-rich device), respectively. Therefore,the used condition is reasonable. This condition works well forgeneral NBIS cases with considering the density of subgaptraps in IGZO and the photoillumination environments incommercial displays.

Based on the tNBIS dependence of AEW, it is found thatwe can calculate the NBIS-induced �VT [lines in Fig. 4(e)]and the stress time evolution of I–V characteristics [linesin Fig. 4(a)–(d)]. It suggests that the proposed calculationmethod can precisely estimate the I–V characteristics as wellas the �VT only from the pristine DOS and NBIS condition,i.e., without long-term stress test.

Fig. 5. Calculated tNBIS dependences of EFn and EC–Eph levels of(a) O-poor and (b) O-rich TFTs. The tNBIS dependences of (c) Teffand (d) AEW.

It is definitely noteworthy that one wants to calculatethe VT and I–V characteristics during NBIS only from thepristine DOS and illumination condition, i.e., without long-term NBIS test. Furthermore, the NBIS-induced �VT canbe described via AEW with the exact function of not onlythe material/process parameters, e.g., gD(E), Eg , gA(E), andNSD, the device geometric parameters, such as the Tact, W , L,and gate insulator thickness, but also the stress conditions,such as the stress gate bias and EFn at bias stress, and theillumination condition, i.e., Eph. The pristine DOS is alsodetermined mainly by the fabrication process and materialnature.

For instance, as shown in Fig. 6, the TFT-B which has arelatively smaller area of the AEW shows a smaller NBIS-induced �VT under specific stress conditions than the �VT

of the TFT-A which has a relatively larger area of the AEWcompared with that of TFT-B. Furthermore, the stress timeevolution of all electrical characteristics can also be calculatedprecisely by using our methodology and parameter set.

JANG et al.: CALCULATION OF NBIS-INDUCED INSTABILITY 1007

Fig. 6. Concepts of instability-aware design and closed loop.

IV. CONCLUSION

Our proposed method, which systematically calculate theNBIS �VT in a-IGZO TFTs, was successfully demonstratedthrough two kinds of specific devices, i.e., O-poor and O-richdevices. Moreover, the effect of oxygen content of a-IGZO onNBIS instability was quantitatively explained very well basedon microscopic origins.

Our results suggest that the tNBIS evolution of I–V charac-teristics can be calculated from the pristine DOS and AEWbefore applying NBIS. Therefore, the proposed method ispotentially useful for calculating/projecting the tNBIS evolutionof I–V characteristics without long-term NBIS test, whichis the essence of instability-aware design. The closed loopin Fig. 6 also suggests that we can both design highly stableoxide TFTs systematically and optimize all the oxide material,fabrication process, and TFT device structure very efficiently.This is the other essence of instability-aware design.

ACKNOWLEDGMENT

The CAD software was supported in part by SILVACO andin part by the IC Design Education Center.

REFERENCES

[1] T. Kamiya, K. Nomura, and H. Hosono, “Present status of amorphousIn–Ga–Zn–O thin-film transistors,” Sci. Technol. Adv. Mater., vol. 11,no. 4, p. 044305, Sep. 2010, doi: 10.1088/1468-6996/11/4/044305.

[2] J. F. Wager, “Flat-panel-display backplanes: LTPS or IGZO forAMLCDs or AMOLED displays?” Inf. Display, vol. 30, no. 2,pp. 26–29, 2014.

[3] J. S. Park, W.-J. Maeng, H.-S. Kim, and J.-S. Park, “Review of recentdevelopments in amorphous oxide semiconductor thin-film transistordevices,” Thin Solid Films, vol. 520, no. 6, pp. 1679–1693, Jan. 2012,doi: 10.1016/j.tsf.2011.07.018.

[4] K. H. Ji et al., “Comparative study on light-induced bias stress instabil-ity of IGZO transistors with SiNx and SiO2 gate dielectrics,” IEEEElectron Device Lett., vol. 31, no. 12, pp. 1404–1406, Dec. 2010,doi: 10.1109/LED.2010.2073439.

[5] J.-Y. Kown et al., “The impact of gate dielectric materials on the light-induced bias instability in Hf–In–Zn–O thin film transistors,” Appl. Phys.Lett., vol. 97, no. 18, p. 183503, Nov. 2010, doi: 10.1063/1.3513400.

[6] A. Janotti and C. G. Van de Walle, “Oxygen vacancies in ZnO,”Appl. Phys. Lett., vol. 87, no. 12, p. 122102, Sep. 2005, doi: 10.1063/1.2053360.

[7] P. Migliorato, M. D. H. Chowdhury, J. G. Um, M. Seok, and J. Jang,“Light/negative bias stress instabilities in indium gallium zinc oxide thinfilm transistors explained by creation of a double donor,” Appl. Phys.Lett., vol. 101, no. 12, p. 123502, Sep. 2012, doi: 10.1063/1.4752238.

[8] H.-H. Nahm, Y.-S. Kim, and D. H. Kim, “Instability of amorphousoxide semiconductors via carrier-mediated structural transition betweendisorder and peroxide state,” Phys. Status Solidi B, vol. 249, no. 6,pp. 1277–1281, Feb. 2012, doi: 10.1002/pssb.201147557.

[9] J. K. Jeong, H. W. Yang, J. H. Jeong, Y.-G. Mo, and H. D. Kim, “Originof threshold voltage instability in indium-gallium-zinc oxide thin filmtransistors,” Appl. Phys. Lett., vol. 93, no. 12, p. 123508, Sep. 2008,doi: 10.1063/1.2990657.

[10] S. Yang et al., “Improvement in the photon-induced bias stability ofAl–Sn–Zn–In–O thin film transistors by adopting AlOx passivationlayer,” Appl. Phys. Lett., vol. 96, no. 21, p. 213511, May 2010,doi: 10.1063/1.3432445.

[11] A. Nathan, S. Lee, S. Jeon, I. Song, and U. I. Chung, “Transparent oxidesemiconductors for advanced display applications,” Inf. Display, vol. 29,no. 1, pp. 6–11, 2013.

[12] Y. W. Jeon et al., “Subgap density-of-states-based amorphous oxide thinfilm transistor simulator (DeAOTS),” IEEE Trans. Electron Devices,vol. 57, no. 11, pp. 2988–3000, Nov. 2010, doi: 10.1109/TED.2010.2072926.

[13] M. Bae, K. M. Lee, E.-S. Cho, H.-I. Kwon, D. M. Kim, andD. H. Kim, “Analytical current and capacitance models for amor-phous indium-gallium-zinc-oxide thin-film transistors,” IEEE Trans.Electron Devices, vol. 60, no. 10, pp. 3465–3473, Oct. 2013,doi: 10.1109/TED.2013.2278033.

[14] H. Bae et al., “Single-scan monochromatic photonic capacitance-voltagetechnique for extraction of subgap DOS over the bandgap in amorphoussemiconductor TFTs,” IEEE Electron Device Lett., vol. 34, no. 12,pp. 1524–1526, Dec. 2013, doi: 10.1109/LED.2013.2287511.

[15] J. T. Jang et al., “Study on the photoresponse of amorphousIn–Ga–Zn–O and zinc oxynitride semiconductor devices by the extrac-tion of sub-gap-state distribution and device simulation,” ACS Appl.Mater. Interfaces, vol. 28, no. 7, pp. 15570–15577, Jun. 2015,doi: 10.1021/acsami.5b04152.

[16] S. Kim et al., “Impact of oxygen flow rate on the instability underpositive bias stresses in dc-sputtered amorphous InGaZnO thin-filmtransistors,” IEEE Electron Device Lett., vol. 33, no. 1, pp. 62–64,Jan. 2012, doi: 10.1109/LED.2011.2173153.

[17] K. H. Ji et al., “Effect of high-pressure oxygen annealing on negativebias illumination stress-induced instability of InGaZnO thin filmtransistors,” Appl. Phys. Lett., vol. 98, no. 10, p. 103509, 2011,doi: 10.1063/1.3564882.

[18] S. Y. Park et al., “Improvement in photo-bias stability of high-mobilityindium zinc oxide thin-film transistors by oxygen high-pressure anneal-ing,” IEEE Electron Device Lett., vol. 34, no. 7, pp. 894–896, Jul. 2013,doi: 10.1109/LED.2013.2259574.

Jun Tae Jang received the B.S. andM.S. degrees in electrical engineering fromKookmin University, Seoul, South Korea, in 2016,where he is currently pursuing the Ph.D. degreewith the Department of Electrical Engineering.

Sung-Jin Choi received the M.S. andPh.D. degrees in electrical engineering fromthe Korea Advanced Institute of Science andTechnology, Daejeon, South Korea, in 2012.

He is currently an Assistant Professor withthe School of Electrical Engineering, KookminUniversity, Seoul, South Korea.

1008 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 65, NO. 3, MARCH 2018

Dong Myong Kim (S’86−M’88) is currently aProfessor with the School of Electrical Engineer-ing, Kookmin University, Seoul, South Korea.His current research interests include fabrica-tion, characterization, modeling of nanostructuresilicon devices, III–V compound semiconductordevices, memories, and CMOS RF circuits.

Dae Hwan Kim (M’08–SM’12) received the B.S.,M.S., and Ph.D. degrees in electrical engineer-ing from Seoul National University, Seoul, SouthKorea, in 1996, 1998, and 2002, respectively.

He is currently a Professor with the Schoolof Electrical Engineering, Kookmin University,Seoul. His current research interests includenano-CMOS, oxide and organic thin-film transis-tors, biosensors, wearable healthcare devices,and emerging memories.